4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
38 #include <sys/types.h>
40 #include <sys/queue.h>
45 #include <sys/param.h>
47 #include <rte_common.h>
48 #include <rte_byteorder.h>
50 #include <rte_memory.h>
51 #include <rte_memcpy.h>
52 #include <rte_memzone.h>
54 #include <rte_per_lcore.h>
55 #include <rte_launch.h>
56 #include <rte_atomic.h>
57 #include <rte_cycles.h>
58 #include <rte_prefetch.h>
59 #include <rte_lcore.h>
60 #include <rte_per_lcore.h>
61 #include <rte_branch_prediction.h>
62 #include <rte_interrupts.h>
64 #include <rte_random.h>
65 #include <rte_debug.h>
66 #include <rte_ether.h>
67 #include <rte_ethdev.h>
69 #include <rte_mempool.h>
71 #include <rte_malloc.h>
75 #include <rte_string_fns.h>
79 #include <rte_ip_frag.h>
81 #define MAX_PKT_BURST 32
84 #define RTE_LOGTYPE_IP_RSMBL RTE_LOGTYPE_USER1
86 #define MAX_JUMBO_PKT_LEN 9600
88 #define BUF_SIZE RTE_MBUF_DEFAULT_DATAROOM
90 (BUF_SIZE + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
94 /* allow max jumbo frame 9.5 KB */
95 #define JUMBO_FRAME_MAX_SIZE 0x2600
97 #define MAX_FLOW_NUM UINT16_MAX
98 #define MIN_FLOW_NUM 1
99 #define DEF_FLOW_NUM 0x1000
101 /* TTL numbers are in ms. */
102 #define MAX_FLOW_TTL (3600 * MS_PER_S)
103 #define MIN_FLOW_TTL 1
104 #define DEF_FLOW_TTL MS_PER_S
106 #define MAX_FRAG_NUM RTE_LIBRTE_IP_FRAG_MAX_FRAG
108 /* Should be power of two. */
109 #define IP_FRAG_TBL_BUCKET_ENTRIES 16
111 static uint32_t max_flow_num = DEF_FLOW_NUM;
112 static uint32_t max_flow_ttl = DEF_FLOW_TTL;
114 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
118 /* Configure how many packets ahead to prefetch, when reading packets */
119 #define PREFETCH_OFFSET 3
122 * Configurable number of RX/TX ring descriptors
124 #define RTE_TEST_RX_DESC_DEFAULT 128
125 #define RTE_TEST_TX_DESC_DEFAULT 512
127 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
128 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
130 /* ethernet addresses of ports */
131 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
134 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8
135 #define IPv4_BYTES(addr) \
136 (uint8_t) (((addr) >> 24) & 0xFF),\
137 (uint8_t) (((addr) >> 16) & 0xFF),\
138 (uint8_t) (((addr) >> 8) & 0xFF),\
139 (uint8_t) ((addr) & 0xFF)
143 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
144 "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
145 #define IPv6_BYTES(addr) \
146 addr[0], addr[1], addr[2], addr[3], \
147 addr[4], addr[5], addr[6], addr[7], \
148 addr[8], addr[9], addr[10], addr[11],\
149 addr[12], addr[13],addr[14], addr[15]
152 #define IPV6_ADDR_LEN 16
154 /* mask of enabled ports */
155 static uint32_t enabled_port_mask = 0;
157 static int rx_queue_per_lcore = 1;
163 struct rte_mbuf *m_table[0];
167 struct rte_ip_frag_tbl *frag_tbl;
168 struct rte_mempool *pool;
170 struct rte_lpm6 *lpm6;
174 struct tx_lcore_stat {
181 #define MAX_RX_QUEUE_PER_LCORE 16
182 #define MAX_TX_QUEUE_PER_PORT 16
183 #define MAX_RX_QUEUE_PER_PORT 128
185 struct lcore_queue_conf {
187 struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
188 uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
189 struct rte_ip_frag_death_row death_row;
190 struct mbuf_table *tx_mbufs[RTE_MAX_ETHPORTS];
191 struct tx_lcore_stat tx_stat;
192 } __rte_cache_aligned;
193 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
195 static struct rte_eth_conf port_conf = {
197 .mq_mode = ETH_MQ_RX_RSS,
198 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
200 .header_split = 0, /**< Header Split disabled */
201 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
202 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
203 .jumbo_frame = 1, /**< Jumbo Frame Support disabled */
204 .hw_strip_crc = 0, /**< CRC stripped by hardware */
209 .rss_hf = ETH_RSS_IP,
213 .mq_mode = ETH_MQ_TX_NONE,
218 * IPv4 forwarding table
220 struct l3fwd_ipv4_route {
226 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = {
227 {IPv4(100,10,0,0), 16, 0},
228 {IPv4(100,20,0,0), 16, 1},
229 {IPv4(100,30,0,0), 16, 2},
230 {IPv4(100,40,0,0), 16, 3},
231 {IPv4(100,50,0,0), 16, 4},
232 {IPv4(100,60,0,0), 16, 5},
233 {IPv4(100,70,0,0), 16, 6},
234 {IPv4(100,80,0,0), 16, 7},
238 * IPv6 forwarding table
241 struct l3fwd_ipv6_route {
242 uint8_t ip[IPV6_ADDR_LEN];
247 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = {
248 {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
249 {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
250 {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
251 {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
252 {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
253 {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
254 {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
255 {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
258 #define LPM_MAX_RULES 1024
259 #define LPM6_MAX_RULES 1024
260 #define LPM6_NUMBER_TBL8S (1 << 16)
262 struct rte_lpm6_config lpm6_config = {
263 .max_rules = LPM6_MAX_RULES,
264 .number_tbl8s = LPM6_NUMBER_TBL8S,
268 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES];
269 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES];
271 #ifdef RTE_LIBRTE_IP_FRAG_TBL_STAT
272 #define TX_LCORE_STAT_UPDATE(s, f, v) ((s)->f += (v))
274 #define TX_LCORE_STAT_UPDATE(s, f, v) do {} while (0)
275 #endif /* RTE_LIBRTE_IP_FRAG_TBL_STAT */
278 * If number of queued packets reached given threahold, then
279 * send burst of packets on an output interface.
281 static inline uint32_t
282 send_burst(struct lcore_queue_conf *qconf, uint32_t thresh, uint8_t port)
284 uint32_t fill, len, k, n;
285 struct mbuf_table *txmb;
287 txmb = qconf->tx_mbufs[port];
290 if ((int32_t)(fill = txmb->head - txmb->tail) < 0)
293 if (fill >= thresh) {
294 n = RTE_MIN(len - txmb->tail, fill);
296 k = rte_eth_tx_burst(port, qconf->tx_queue_id[port],
297 txmb->m_table + txmb->tail, (uint16_t)n);
299 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, call, 1);
300 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, send, k);
303 if ((txmb->tail += k) == len)
310 /* Enqueue a single packet, and send burst if queue is filled */
312 send_single_packet(struct rte_mbuf *m, uint8_t port)
314 uint32_t fill, lcore_id, len;
315 struct lcore_queue_conf *qconf;
316 struct mbuf_table *txmb;
318 lcore_id = rte_lcore_id();
319 qconf = &lcore_queue_conf[lcore_id];
321 txmb = qconf->tx_mbufs[port];
324 fill = send_burst(qconf, MAX_PKT_BURST, port);
326 if (fill == len - 1) {
327 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, drop, 1);
328 rte_pktmbuf_free(txmb->m_table[txmb->tail]);
329 if (++txmb->tail == len)
333 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, queue, 1);
334 txmb->m_table[txmb->head] = m;
335 if(++txmb->head == len)
342 reassemble(struct rte_mbuf *m, uint8_t portid, uint32_t queue,
343 struct lcore_queue_conf *qconf, uint64_t tms)
345 struct ether_hdr *eth_hdr;
346 struct rte_ip_frag_tbl *tbl;
347 struct rte_ip_frag_death_row *dr;
348 struct rx_queue *rxq;
350 uint8_t next_hop, dst_port;
352 rxq = &qconf->rx_queue_list[queue];
354 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
358 /* if packet is IPv4 */
359 if (m->ol_flags & (PKT_RX_IPV4_HDR)) {
360 struct ipv4_hdr *ip_hdr;
363 ip_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
365 /* if it is a fragmented packet, then try to reassemble. */
366 if (rte_ipv4_frag_pkt_is_fragmented(ip_hdr)) {
370 dr = &qconf->death_row;
372 /* prepare mbuf: setup l2_len/l3_len. */
373 m->l2_len = sizeof(*eth_hdr);
374 m->l3_len = sizeof(*ip_hdr);
376 /* process this fragment. */
377 mo = rte_ipv4_frag_reassemble_packet(tbl, dr, m, tms, ip_hdr);
379 /* no packet to send out. */
382 /* we have our packet reassembled. */
385 eth_hdr = rte_pktmbuf_mtod(m,
387 ip_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
390 ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
392 /* Find destination port */
393 if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop) == 0 &&
394 (enabled_port_mask & 1 << next_hop) != 0) {
398 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
400 /* if packet is IPv6 */
401 else if (m->ol_flags & (PKT_RX_IPV6_HDR | PKT_RX_IPV6_HDR_EXT)) {
402 struct ipv6_extension_fragment *frag_hdr;
403 struct ipv6_hdr *ip_hdr;
405 ip_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
407 frag_hdr = rte_ipv6_frag_get_ipv6_fragment_header(ip_hdr);
409 if (frag_hdr != NULL) {
413 dr = &qconf->death_row;
415 /* prepare mbuf: setup l2_len/l3_len. */
416 m->l2_len = sizeof(*eth_hdr);
417 m->l3_len = sizeof(*ip_hdr) + sizeof(*frag_hdr);
419 mo = rte_ipv6_frag_reassemble_packet(tbl, dr, m, tms, ip_hdr, frag_hdr);
425 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
426 ip_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
430 /* Find destination port */
431 if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, &next_hop) == 0 &&
432 (enabled_port_mask & 1 << next_hop) != 0) {
436 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6);
438 /* if packet wasn't IPv4 or IPv6, it's forwarded to the port it came from */
440 /* 02:00:00:00:00:xx */
441 d_addr_bytes = ð_hdr->d_addr.addr_bytes[0];
442 *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)dst_port << 40);
445 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr);
447 send_single_packet(m, dst_port);
450 /* main processing loop */
452 main_loop(__attribute__((unused)) void *dummy)
454 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
456 uint64_t diff_tsc, cur_tsc, prev_tsc;
459 struct lcore_queue_conf *qconf;
460 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
464 lcore_id = rte_lcore_id();
465 qconf = &lcore_queue_conf[lcore_id];
467 if (qconf->n_rx_queue == 0) {
468 RTE_LOG(INFO, IP_RSMBL, "lcore %u has nothing to do\n", lcore_id);
472 RTE_LOG(INFO, IP_RSMBL, "entering main loop on lcore %u\n", lcore_id);
474 for (i = 0; i < qconf->n_rx_queue; i++) {
476 portid = qconf->rx_queue_list[i].portid;
477 RTE_LOG(INFO, IP_RSMBL, " -- lcoreid=%u portid=%hhu\n", lcore_id,
483 cur_tsc = rte_rdtsc();
486 * TX burst queue drain
488 diff_tsc = cur_tsc - prev_tsc;
489 if (unlikely(diff_tsc > drain_tsc)) {
492 * This could be optimized (use queueid instead of
493 * portid), but it is not called so often
495 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
496 if ((enabled_port_mask & (1 << portid)) != 0)
497 send_burst(qconf, 1, portid);
504 * Read packet from RX queues
506 for (i = 0; i < qconf->n_rx_queue; ++i) {
508 portid = qconf->rx_queue_list[i].portid;
510 nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
513 /* Prefetch first packets */
514 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
515 rte_prefetch0(rte_pktmbuf_mtod(
516 pkts_burst[j], void *));
519 /* Prefetch and forward already prefetched packets */
520 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
521 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
522 j + PREFETCH_OFFSET], void *));
523 reassemble(pkts_burst[j], portid,
527 /* Forward remaining prefetched packets */
528 for (; j < nb_rx; j++) {
529 reassemble(pkts_burst[j], portid,
533 rte_ip_frag_free_death_row(&qconf->death_row,
541 print_usage(const char *prgname)
543 printf("%s [EAL options] -- -p PORTMASK [-q NQ]"
544 " [--max-pkt-len PKTLEN]"
545 " [--maxflows=<flows>] [--flowttl=<ttl>[(s|ms)]]\n"
546 " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
547 " -q NQ: number of RX queues per lcore\n"
548 " --maxflows=<flows>: optional, maximum number of flows "
550 " --flowttl=<ttl>[(s|ms)]: optional, maximum TTL for each "
556 parse_flow_num(const char *str, uint32_t min, uint32_t max, uint32_t *val)
561 /* parse decimal string */
563 v = strtoul(str, &end, 10);
564 if (errno != 0 || *end != '\0')
567 if (v < min || v > max)
575 parse_flow_ttl(const char *str, uint32_t min, uint32_t max, uint32_t *val)
580 static const char frmt_sec[] = "s";
581 static const char frmt_msec[] = "ms";
583 /* parse decimal string */
585 v = strtoul(str, &end, 10);
590 if (strncmp(frmt_sec, end, sizeof(frmt_sec)) == 0)
592 else if (strncmp(frmt_msec, end, sizeof (frmt_msec)) != 0)
596 if (v < min || v > max)
604 parse_portmask(const char *portmask)
609 /* parse hexadecimal string */
610 pm = strtoul(portmask, &end, 16);
611 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
621 parse_nqueue(const char *q_arg)
626 printf("%p\n", q_arg);
628 /* parse hexadecimal string */
629 n = strtoul(q_arg, &end, 10);
630 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
634 if (n >= MAX_RX_QUEUE_PER_LCORE)
640 /* Parse the argument given in the command line of the application */
642 parse_args(int argc, char **argv)
647 char *prgname = argv[0];
648 static struct option lgopts[] = {
649 {"max-pkt-len", 1, 0, 0},
650 {"maxflows", 1, 0, 0},
651 {"flowttl", 1, 0, 0},
657 while ((opt = getopt_long(argc, argvopt, "p:q:",
658 lgopts, &option_index)) != EOF) {
663 enabled_port_mask = parse_portmask(optarg);
664 if (enabled_port_mask == 0) {
665 printf("invalid portmask\n");
666 print_usage(prgname);
673 rx_queue_per_lcore = parse_nqueue(optarg);
674 if (rx_queue_per_lcore < 0) {
675 printf("invalid queue number\n");
676 print_usage(prgname);
683 if (!strncmp(lgopts[option_index].name,
685 if ((ret = parse_flow_num(optarg, MIN_FLOW_NUM,
687 &max_flow_num)) != 0) {
688 printf("invalid value: \"%s\" for "
691 lgopts[option_index].name);
692 print_usage(prgname);
697 if (!strncmp(lgopts[option_index].name, "flowttl", 7)) {
698 if ((ret = parse_flow_ttl(optarg, MIN_FLOW_TTL,
700 &max_flow_ttl)) != 0) {
701 printf("invalid value: \"%s\" for "
704 lgopts[option_index].name);
705 print_usage(prgname);
713 print_usage(prgname);
719 argv[optind-1] = prgname;
722 optind = 0; /* reset getopt lib */
727 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
729 char buf[ETHER_ADDR_FMT_SIZE];
730 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
731 printf("%s%s", name, buf);
734 /* Check the link status of all ports in up to 9s, and print them finally */
736 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
738 #define CHECK_INTERVAL 100 /* 100ms */
739 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
740 uint8_t portid, count, all_ports_up, print_flag = 0;
741 struct rte_eth_link link;
743 printf("\nChecking link status");
745 for (count = 0; count <= MAX_CHECK_TIME; count++) {
747 for (portid = 0; portid < port_num; portid++) {
748 if ((port_mask & (1 << portid)) == 0)
750 memset(&link, 0, sizeof(link));
751 rte_eth_link_get_nowait(portid, &link);
752 /* print link status if flag set */
753 if (print_flag == 1) {
754 if (link.link_status)
755 printf("Port %d Link Up - speed %u "
756 "Mbps - %s\n", (uint8_t)portid,
757 (unsigned)link.link_speed,
758 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
759 ("full-duplex") : ("half-duplex\n"));
761 printf("Port %d Link Down\n",
765 /* clear all_ports_up flag if any link down */
766 if (link.link_status == 0) {
771 /* after finally printing all link status, get out */
775 if (all_ports_up == 0) {
778 rte_delay_ms(CHECK_INTERVAL);
781 /* set the print_flag if all ports up or timeout */
782 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
790 init_routing_table(void)
793 struct rte_lpm6 *lpm6;
797 for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
798 if (socket_lpm[socket]) {
799 lpm = socket_lpm[socket];
800 /* populate the LPM table */
801 for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) {
802 ret = rte_lpm_add(lpm,
803 l3fwd_ipv4_route_array[i].ip,
804 l3fwd_ipv4_route_array[i].depth,
805 l3fwd_ipv4_route_array[i].if_out);
808 RTE_LOG(ERR, IP_RSMBL, "Unable to add entry %i to the l3fwd "
813 RTE_LOG(INFO, IP_RSMBL, "Socket %i: adding route " IPv4_BYTES_FMT
816 IPv4_BYTES(l3fwd_ipv4_route_array[i].ip),
817 l3fwd_ipv4_route_array[i].depth,
818 l3fwd_ipv4_route_array[i].if_out);
822 if (socket_lpm6[socket]) {
823 lpm6 = socket_lpm6[socket];
824 /* populate the LPM6 table */
825 for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) {
826 ret = rte_lpm6_add(lpm6,
827 l3fwd_ipv6_route_array[i].ip,
828 l3fwd_ipv6_route_array[i].depth,
829 l3fwd_ipv6_route_array[i].if_out);
832 RTE_LOG(ERR, IP_RSMBL, "Unable to add entry %i to the l3fwd "
837 RTE_LOG(INFO, IP_RSMBL, "Socket %i: adding route " IPv6_BYTES_FMT
840 IPv6_BYTES(l3fwd_ipv6_route_array[i].ip),
841 l3fwd_ipv6_route_array[i].depth,
842 l3fwd_ipv6_route_array[i].if_out);
850 setup_port_tbl(struct lcore_queue_conf *qconf, uint32_t lcore, int socket,
853 struct mbuf_table *mtb;
857 n = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST);
858 sz = sizeof (*mtb) + sizeof (mtb->m_table[0]) * n;
860 if ((mtb = rte_zmalloc_socket(__func__, sz, RTE_CACHE_LINE_SIZE,
862 RTE_LOG(ERR, IP_RSMBL, "%s() for lcore: %u, port: %u "
863 "failed to allocate %zu bytes\n",
864 __func__, lcore, port, sz);
869 qconf->tx_mbufs[port] = mtb;
875 setup_queue_tbl(struct rx_queue *rxq, uint32_t lcore, uint32_t queue)
879 uint64_t frag_cycles;
880 char buf[RTE_MEMPOOL_NAMESIZE];
882 socket = rte_lcore_to_socket_id(lcore);
883 if (socket == SOCKET_ID_ANY)
886 frag_cycles = (rte_get_tsc_hz() + MS_PER_S - 1) / MS_PER_S *
889 if ((rxq->frag_tbl = rte_ip_frag_table_create(max_flow_num,
890 IP_FRAG_TBL_BUCKET_ENTRIES, max_flow_num, frag_cycles,
892 RTE_LOG(ERR, IP_RSMBL, "ip_frag_tbl_create(%u) on "
893 "lcore: %u for queue: %u failed\n",
894 max_flow_num, lcore, queue);
899 * At any given moment up to <max_flow_num * (MAX_FRAG_NUM)>
900 * mbufs could be stored int the fragment table.
901 * Plus, each TX queue can hold up to <max_flow_num> packets.
904 nb_mbuf = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST) * MAX_FRAG_NUM;
905 nb_mbuf *= (port_conf.rxmode.max_rx_pkt_len + BUF_SIZE - 1) / BUF_SIZE;
906 nb_mbuf *= 2; /* ipv4 and ipv6 */
907 nb_mbuf += RTE_TEST_RX_DESC_DEFAULT + RTE_TEST_TX_DESC_DEFAULT;
909 nb_mbuf = RTE_MAX(nb_mbuf, (uint32_t)NB_MBUF);
911 snprintf(buf, sizeof(buf), "mbuf_pool_%u_%u", lcore, queue);
913 if ((rxq->pool = rte_mempool_create(buf, nb_mbuf, MBUF_SIZE, 0,
914 sizeof(struct rte_pktmbuf_pool_private),
915 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
916 socket, MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET)) == NULL) {
917 RTE_LOG(ERR, IP_RSMBL, "mempool_create(%s) failed", buf);
929 struct rte_lpm6 *lpm6;
933 /* traverse through lcores and initialize structures on each socket */
935 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
937 if (rte_lcore_is_enabled(lcore_id) == 0)
940 socket = rte_lcore_to_socket_id(lcore_id);
942 if (socket == SOCKET_ID_ANY)
945 if (socket_lpm[socket] == NULL) {
946 RTE_LOG(INFO, IP_RSMBL, "Creating LPM table on socket %i\n", socket);
947 snprintf(buf, sizeof(buf), "IP_RSMBL_LPM_%i", socket);
949 lpm = rte_lpm_create(buf, socket, LPM_MAX_RULES, 0);
951 RTE_LOG(ERR, IP_RSMBL, "Cannot create LPM table\n");
954 socket_lpm[socket] = lpm;
957 if (socket_lpm6[socket] == NULL) {
958 RTE_LOG(INFO, IP_RSMBL, "Creating LPM6 table on socket %i\n", socket);
959 snprintf(buf, sizeof(buf), "IP_RSMBL_LPM_%i", socket);
961 lpm6 = rte_lpm6_create("IP_RSMBL_LPM6", socket, &lpm6_config);
963 RTE_LOG(ERR, IP_RSMBL, "Cannot create LPM table\n");
966 socket_lpm6[socket] = lpm6;
974 queue_dump_stat(void)
977 const struct lcore_queue_conf *qconf;
979 for (lcore = 0; lcore < RTE_MAX_LCORE; lcore++) {
980 if (rte_lcore_is_enabled(lcore) == 0)
983 qconf = &lcore_queue_conf[lcore];
984 for (i = 0; i < qconf->n_rx_queue; i++) {
986 fprintf(stdout, " -- lcoreid=%u portid=%hhu "
988 lcore, qconf->rx_queue_list[i].portid);
989 rte_ip_frag_table_statistics_dump(stdout,
990 qconf->rx_queue_list[i].frag_tbl);
991 fprintf(stdout, "TX bursts:\t%" PRIu64 "\n"
992 "TX packets _queued:\t%" PRIu64 "\n"
993 "TX packets dropped:\t%" PRIu64 "\n"
994 "TX packets send:\t%" PRIu64 "\n",
996 qconf->tx_stat.queue,
998 qconf->tx_stat.send);
1004 signal_handler(int signum)
1007 if (signum != SIGUSR1)
1008 rte_exit(0, "received signal: %d, exiting\n", signum);
1012 main(int argc, char **argv)
1014 struct lcore_queue_conf *qconf;
1015 struct rte_eth_dev_info dev_info;
1016 struct rte_eth_txconf *txconf;
1017 struct rx_queue *rxq;
1021 unsigned lcore_id = 0, rx_lcore_id = 0;
1022 uint32_t n_tx_queue, nb_lcores;
1026 ret = rte_eal_init(argc, argv);
1028 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1032 /* parse application arguments (after the EAL ones) */
1033 ret = parse_args(argc, argv);
1035 rte_exit(EXIT_FAILURE, "Invalid IP reassembly parameters\n");
1037 nb_ports = rte_eth_dev_count();
1038 if (nb_ports > RTE_MAX_ETHPORTS)
1039 nb_ports = RTE_MAX_ETHPORTS;
1040 else if (nb_ports == 0)
1041 rte_exit(EXIT_FAILURE, "No ports found!\n");
1043 nb_lcores = rte_lcore_count();
1045 /* initialize structures (mempools, lpm etc.) */
1047 rte_panic("Cannot initialize memory structures!\n");
1049 /* check if portmask has non-existent ports */
1050 if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned)))
1051 rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n");
1053 /* initialize all ports */
1054 for (portid = 0; portid < nb_ports; portid++) {
1055 /* skip ports that are not enabled */
1056 if ((enabled_port_mask & (1 << portid)) == 0) {
1057 printf("\nSkipping disabled port %d\n", portid);
1061 qconf = &lcore_queue_conf[rx_lcore_id];
1063 /* get the lcore_id for this port */
1064 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
1065 qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
1068 if (rx_lcore_id >= RTE_MAX_LCORE)
1069 rte_exit(EXIT_FAILURE, "Not enough cores\n");
1071 qconf = &lcore_queue_conf[rx_lcore_id];
1074 socket = rte_lcore_to_socket_id(portid);
1075 if (socket == SOCKET_ID_ANY)
1078 queueid = qconf->n_rx_queue;
1079 rxq = &qconf->rx_queue_list[queueid];
1080 rxq->portid = portid;
1081 rxq->lpm = socket_lpm[socket];
1082 rxq->lpm6 = socket_lpm6[socket];
1083 if (setup_queue_tbl(rxq, rx_lcore_id, queueid) < 0)
1084 rte_exit(EXIT_FAILURE, "Failed to set up queue table\n");
1085 qconf->n_rx_queue++;
1088 printf("Initializing port %d ... ", portid );
1091 n_tx_queue = nb_lcores;
1092 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1093 n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1094 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
1098 rte_exit(EXIT_FAILURE, "Cannot configure device: "
1099 "err=%d, port=%d\n",
1103 /* init one RX queue */
1104 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
1109 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
1110 "err=%d, port=%d\n",
1114 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1115 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1118 /* init one TX queue per couple (lcore,port) */
1120 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1121 if (rte_lcore_is_enabled(lcore_id) == 0)
1124 socket = (int) rte_lcore_to_socket_id(lcore_id);
1126 printf("txq=%u,%d,%d ", lcore_id, queueid, socket);
1129 rte_eth_dev_info_get(portid, &dev_info);
1130 txconf = &dev_info.default_txconf;
1131 txconf->txq_flags = 0;
1133 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
1136 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
1137 "port=%d\n", ret, portid);
1139 qconf = &lcore_queue_conf[lcore_id];
1140 qconf->tx_queue_id[portid] = queueid;
1141 setup_port_tbl(qconf, lcore_id, socket, portid);
1150 for (portid = 0; portid < nb_ports; portid++) {
1151 if ((enabled_port_mask & (1 << portid)) == 0) {
1155 ret = rte_eth_dev_start(portid);
1157 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1160 rte_eth_promiscuous_enable(portid);
1163 if (init_routing_table() < 0)
1164 rte_exit(EXIT_FAILURE, "Cannot init routing table\n");
1166 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
1168 signal(SIGUSR1, signal_handler);
1169 signal(SIGTERM, signal_handler);
1170 signal(SIGINT, signal_handler);
1172 /* launch per-lcore init on every lcore */
1173 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1174 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1175 if (rte_eal_wait_lcore(lcore_id) < 0)